Data Domain doubles VTL performance

By Kevin Komiega

—Data Domain plans to beef up its high-end virtual tape libraries (VTLs) with a faster, more-scalable de-duplication system that offers twice the performance of previous systems.

The new DD580, along with the DD580g gateway version, leverages Data Domain's Stream-Informed Segment Layout (SISL) architecture and takes advantage of multi-core processors to make it possible for a single DD580 to deliver up to 800GB/hour of aggregate de-duplication throughput with less than 15 SATA RAID-6-protected disks.

A fully configured DDX Array with 16 DD580 controllers increases throughput performance to more than 12TB/hour and offers up to 20PB of capacity for long-term data retention.

The DD580 is available as an appliance, or as the DD580g gateway, which supports external Fibre Channel or SATA disk arrays. Both versions support existing backup software products on any type of storage fabric.

"Our systems really serve as a target for backup software, and they're fabric-independent so they can look like a NAS filer or VTL," says Brian Biles, co-founder and vice president of product management at Data Domain. "The systems offer both a file system and VTL interface on the same box."

The DD580 can automate WAN vaulting for use in remote-office disaster-recovery scenarios, or for multi-site tape consolidation. The systems de-duplicate data inline before storing to disk and replicate de-duplicated data while the backup is in progress, versus the post-process approach employed by some of Data Domain's competitors.

"We're narrowing the performance gap between inline and post-process de-duplication to the point where it may not matter anymore. The aggregate speed per controller is getting pretty close," says Biles. "When it gets this close users have to ask why they need so many disks and begin looking at more-scalable architectures. Some other [VTL] vendors are disk-bound while we're CPU-centric."

Vendors that offer de-duplication have been at odds for some time over which approach is best?inline or post-process. Heidi Biggar, an analyst with the Enterprise Strategy Group, says the debate over inline versus post-process de-duplication has taken on a life of its own.

"To put the inline versus post-process debate in perspective, they are both good technologies. Data de-duplication changes the economics of disk-based data protection. Period. But users should ask a series of questions to ensure they choose the right de-duplication solution for their environments," says Biggar.

The questions users should be asking, she says, relate to capacity versus performance and the trade-offs associated with each approach.

"Beyond de-duplication speed, the length of the overall data-protection process has to be considered?that is, from start to finish: How long does it take for the backup, de-duplication, and disaster-recovery process to complete? Users ultimately need to weigh de-duplication technologies holistically," says Biggar. "All that really matters is when the process is completed, and how this maps to your service level agreements."

Biggar says Data Domain's CPU-centric approach toward scaling performance is a potentially more cost-effective way to scale versus adding more disk spindles to build out the environment, but it does not completely eliminate the performance impacts of inline de-duplication.

So far, Data Domain has focused primarily on the midrange market, but the performance boost offered by the new DD580 might help the company gain traction with larger enterprises. "Increasing the performance potentially opens new doors in the enterprise space," says Biggar. "Data Domain is in a good position to give competitors a run for their money in the enterprise."

The DD580 will be generally available this summer. A base configuration of the appliance with 15 drives is priced from $120,000.

This article was originally published on May 11, 2007